Accelerating flat-histogram methods for potential of mean force calculations

Area of Science:

• Computational chemistry and biophysics.
• Statistical mechanics and molecular simulations.

Background:

• Calculating potential of mean force (PMF) requires extensive sampling, leading to long computational times.
• The expanded ensemble density of states (EXEDOS) method offers uniform sampling but faces convergence and tunneling time limitations.

Purpose of the Study:

• To accelerate EXEDOS for more efficient free energy calculations.
• To overcome limitations of EXEDOS in terms of convergence and sampling.

Main Methods:

• Developed an asynchronous parallel implementation of the density of states algorithm in a multiple-walkers multiple-windows scheme, extended to an expanded ensemble [(MW)(2)-XDOS].
• Coupled (MW)(2)-XDOS with rejection-free geometric cluster moves for low-density systems.
• Integrated (MW)(2)-XDOS with preferential sampling methods, including single and collective particle displacements ('local Brownian dynamics'), for high-density and complex molecular systems.

Main Results:

• The parallelized (MW)(2)-XDOS overcomes limitations related to extreme tunneling times.
• Geometric cluster moves combined with (MW)(2)-XDOS show superior performance at low densities.
• Preferential sampling significantly accelerates calculations and achieves near size-independence, especially at high densities and for complex systems.

Conclusions:

• The proposed improvements substantially accelerate EXEDOS, making PMF calculations more feasible.
• The developed methods are broadly applicable to various free energy simulation techniques.
• These advancements offer significant benefits for high-density and molecular systems with strong potentials.